Prediction of limiting activity coefficients for binary vapor-liquid equilibrium using neural networks

Behrooz, Hesam Ahmadian; Boozarjomehry, Ramin Bozorgmehry

doi:10.1016/j.fluid.2016.10.033

Cited by 23 publications

(8 citation statements)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Techniques such as gas–liquid chromatography, , high-performance liquid chromatography, and differential ebulliometry ,, are traditionally used to measure activities in extremely dilute systems at varying concentrations, leading to the IDAC by extrapolation . There has been considerable interest in predicting these coefficients − ,,− due to their use in phase equilibria studies in chemical engineering applications. ,,, …”

Section: Methodsmentioning

confidence: 99%

Infinite Dilution Activity Coefficients as Constraints for Force Field Parametrization and Method Development

Matos

Calabró²,

Mobley

2019

J. Chem. Theory Comput.

View full text Add to dashboard Cite

Molecular simulations begin with an underlying energy model or force field, and from this, can predict diverse physical properties. However, force fields were often developed with relatively limited datasets, yet accuracy for diverse properties across a broad chemical space is desirable, so tests of such accuracy are particularly important. Here, to this end, we calculate 237 infinite dilution activity coefficients (IDACs), comparing with experimental values from NIST’s ThermoML database. We found that calculated IDAC values correlate strongly with experiment (Pearson R of 0.92±0.01), and allow us to identify specific functional groups which appear to present challenges to the force field employed. One potentially valuable aspect of IDACs, as compared to solvation free energies which have been frequently employed as force field tests, is that the same molecules serve both as solutes and solvents in different cases, allowing us to ensure force fields are not overly tuned to one particular environment or solvent.

show abstract

Section: Methodsmentioning

confidence: 99%

Infinite Dilution Activity Coefficients as Constraints for Force Field Parametrization and Method Development

Matos

Calabró²,

Mobley

2019

J. Chem. Theory Comput.

View full text Add to dashboard Cite

show abstract

“…18,005 available data points, including 235 solutes, 352 solvents, and 6680 solute/solvent combinations at temperatures ranging from 253.15 to 555.6 K, are obtained. 6441 γ ∞ for 132 compounds are collected from the literature used for revising MOSCED parameters 16 Over 1800 γ ∞ in temperature range of 249.84–412.60 K are gathered from the literature utilized for developing an artificial neural network model 24 411 γ ∞ at 298.15 K for 168 components are collected from DECHEMA for developing a neural network‐based QSPR model 25 …”

Section: Dataset Formulationmentioning

confidence: 99%

“…It was not until this century that machine learning was employed in the prediction of γ ∞ . Behrooz and Boozarjomehry gathered more than 1800 γ ∞ at different temperatures from open literature and used them to develop an artificial neural network‐based method for γ ∞ prediction 24 . Molecular weight, dipole moment, critical temperature, critical pressure, critical volume, and acentric factor for both solutes and solvents, along with the system temperature, were selected as model inputs.…”

Section: Introductionmentioning

confidence: 99%

Prediction of infinite‐dilution activity coefficients with neural collaborative filtering

et al. 2022

View full text Add to dashboard Cite

Accurate prediction of infinite dilution activity coefficient (γ ∞ ) for phase equilibria and process design is crucial. In this work, an experimental γ ∞ dataset containing 295 solutes and 407 solvents (21,048 points) is obtained through data integrating, cleaning, and filtering. The dataset is arranged as a sparse matrix with solutes and solvents as columns and rows, respectively. Neural collaborative filtering (NCF), a modern matrix completion technique based on deep learning, is proposed to fully fill in the γ ∞ matrix. Ten-fold cross-validation is performed on the collected dataset to test the effectiveness of the proposed NCF, proving that NCF outperforms the state-ofthe-art physical model and previous machine learning model. The completed γ ∞ matrix makes solvent screening and extension of UNIFAC parameters possible. Taking two typical hard-to-separate systems (benzene/cyclohexane and methyl cyclopentane/n-hexane mixtures) as examples, the NCF-developed database provides high-throughput screening for separation systems in terms of solvent selectivity and capacity.

show abstract

“…At the opposite side of the methods spectrum are machine learning (ML) methods that have also been used to predict activity coefficients, solvation free energies, or other fluid properties for large data sets, while being difficult to interpret in physical terms. − To make use of the advantages of both, physically based and ML approaches, a combination has been suggested that is based on fingerprints generated from short equilibrium MD simulations that encompass information on the distributions of potential energy of the solute (and its LJ and electrostatic components), radius of gyration, and solvent-accessible surface area . This approach was used successfully to predict solvation free energies in five different solvents as well as partition coefficients in three solvent combinations .…”

Section: Introductionmentioning

confidence: 99%

Combining Molecular Dynamics and Machine Learning to Predict Self-Solvation Free Energies and Limiting Activity Coefficients

Gebhardt

Kiesel

Riniker

et al. 2020

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Computational prediction of limiting activity coefficients is of great relevance for process design. For highly nonideal mixtures including molecules with directed interactions, methods that maintain the molecular character of the solvent are most promising. Computational expense and force-field deficiencies are the main limiting factors that prevent the use of highthroughput molecular dynamics (MD) simulations in a predictive setup. The combination of MD simulations and machine learning used in this work accounts for both issues. Comparison to published data including free-energy simulations, COSMO-RS and UNIFAC models, reveals competitive prediction accuracy.

show abstract

Prediction of limiting activity coefficients for binary vapor-liquid equilibrium using neural networks

Cited by 23 publications

References 65 publications

Infinite Dilution Activity Coefficients as Constraints for Force Field Parametrization and Method Development

Infinite Dilution Activity Coefficients as Constraints for Force Field Parametrization and Method Development

Prediction of infinite‐dilution activity coefficients with neural collaborative filtering

Combining Molecular Dynamics and Machine Learning to Predict Self-Solvation Free Energies and Limiting Activity Coefficients

Contact Info

Product

Resources

About